There were three recognized influenza pandemics in the 20th century, and historical experience suggests that another influenza pandemic is all but inevitable. Although 'colds' and seasonal influenza are clinically very different diseases from pandemic influenza, they share common transmission pathways and the community-level interventions needed to reduce both seasonal flu, common viral URIs and pandemic influenza are likely to be similar. In addition, the circulating influenza viruses causing seasonal flu are physically identical to pandemic influenza viruses, and therefore should be a good model for testing interventions before a pandemic emerges. Little research has been conducted to identify influenza prevention strategies at the household level. Hence, specific aims of this project are to compare the impact of two household-level interventions (an alcohol-based hand sanitizer with or without face masks) on six outcomes: incidence and types or strains of virologically confirmed influenza occurring in study households; rates of symptoms of influenza and viral URIs; basic reproduction number (R0), i.e. number of secondary cases generated by a single infected person in a fully susceptible household; antibiotic use practices for symptoms of influenza and other viral URIs; household member knowledge of prevention and treatment strategies for pandemic influenza and viral URIs; and rates of influenza vaccination among household members. 450 households in northern Manhattan (primarily recently immigrated Hispanics) will be randomized to three groups: control (receiving only a pamphlet on influenza prevention), alcohol hand sanitizer, and sanitizer plus face masks. Symptoms of influenza will be monitored daily for 15 months using ecological momentary assessment technology. Virologic cultures will be obtained from persons with flu symptoms (fever >100 degrees F., sore throat and/or cough). Antibiotic use practices, knowledge, and vaccination rates will be assessed by survey using pilot-tested, psychometrically sound instruments. For this cluster randomization design with randomized intervention on the household level, outcomes will be measured at the individual and household level using generalized linear mixed model for counts response with a Poisson distribution and other appropriate multivariate techniques to control for confounding.